Predicting a behavior of a road used using one or more coarse contextual information

Abstract

A method for predicting behaviors of road users, the method may include sensing a vicinity of a vehicle to provide sensed information; processing the sensed information to provide compact contextual signatures of sensed road users within the vicinity of the vehicle; wherein a compact contextual signature of each a sensed road user includes (a) coarse contextual metadata regarding the sensed road user, (b) coarse location information regarding the sensed road user, (c) identifiers of other sensed road users, and (d) coarse situation information; feeding the compact contextual signatures to a machine learning process trained to estimate behaviors of road users based on compact contextual signatures of road users; and predicting, by the machine learning process, the behaviors of the sensed road users.

What is claimed is: 
     
       1. A method for predicting behaviors of road users, the method comprises:
 sensing a vicinity of a vehicle to provide sensed information; 
 processing the sensed information to provide compact contextual signatures of sensed road users within the vicinity of the vehicle; wherein a compact contextual signature of each of a sensed road user comprises (a) coarse contextual metadata regarding the sensed road user, (b) coarse location information regarding the sensed road user, (c) identifiers of other sensed road users, and (d) coarse situation information; 
 feeding the compact contextual signatures to a machine learning process trained to estimate behaviors of road users based on the compact contextual signatures of road users; and 
 predicting, by the machine learning process, the behaviors of the sensed road users. 
 
     
     
       2. The method according to  claim 1  wherein the compact contextual signature of each sensed road user consists essentially of (a) the coarse contextual metadata regarding the sensed road user, (b) the coarse location information regarding the sensed road user, (c) the identifiers of other sensed road users, and (d) the coarse situation information. 
     
     
       3. The method according to  claim 2  wherein the coarse location information consists essentially of (a) a segment in which the road user is located, and (b) location of the road user within the segment. 
     
     
       4. The method according to  claim 3  wherein the location information reflects a location of the road user within the segment during a period that exceeds one second. 
     
     
       5. The method according to  claim 2  wherein the coarse contextual metadata regarding the sensed road user consists essentially of (a) a type of the road user, and (b) one or more motion related attributes. 
     
     
       6. The method according to  claim 5  wherein the one or more motion related attribute consists essentially of a movement indicator of the road user. 
     
     
       7. The method according to  claim 2  wherein the coarse situation information comprises environmental metadata that illustrates segments of the environment. 
     
     
       8. The method according to  claim 7  wherein the environmental information consists essentially of (a) segments coarse dimensional information, (b) segments orientation, (c) legal limitations information, and (d) exit information regarding allowable exit directions from segments. 
     
     
       9. A non-transitory computer readable medium that stores instructions that, when executed, cause a processor to:
 sensing a vicinity of a vehicle to provide sensed information; 
 processing the sensed information to provide compact contextual signatures of sensed road users within the vicinity of the vehicle; wherein a compact contextual signature of each of a sensed road user comprises (a) coarse contextual metadata regarding the sensed road user, (b) coarse location information regarding the sensed road user, (c) identifiers of other sensed road users, and (d) coarse situation information; 
 feeding the compact contextual signatures to a machine learning process trained to estimate behaviors of road users based on the compact contextual signatures of road users; and 
 predicting, by the machine learning process, the behaviors of the sensed road users. 
 
     
     
       10. The non-transitory computer readable medium according to  claim 9  wherein the compact contextual signature of each sensed road user consists essentially of (a) the coarse contextual metadata regarding the sensed road user, (b) the coarse location information regarding the sensed road user, (c) the identifiers of other sensed road users, and (d) the coarse situation information. 
     
     
       11. The non-transitory computer readable medium according to  claim 10  wherein the coarse location information consists essentially of (a) a segment in which the road user is located, and (b) location of the road user within the segment. 
     
     
       12. The non-transitory computer readable medium according to  claim 11  wherein the location information reflects a location of the road user within the segment during a period that exceeds one second. 
     
     
       13. The non-transitory computer readable medium according to  claim 10  wherein the coarse contextual metadata regarding the sensed road user consists essentially of (a) a type of the road user, and (b) one or more motion related attributes. 
     
     
       14. The non-transitory computer readable medium according to  claim 13  wherein the one or more motion related attribute consists essentially of a movement indicator of the road user. 
     
     
       15. The non-transitory computer readable medium according to  claim 10  wherein the coarse situation information comprises environmental metadata that illustrates segments of the environment. 
     
     
       16. The non-transitory computer readable medium according to  claim 15  wherein the environmental information consists essentially of (a) segments coarse dimensional information, (b) segments orientation, (c) legal limitations information, and (d) exit information regarding allowable exit directions from segments. 
     
     
       17. A computerized system for predicting behaviors of road users, the computerized system comprises:
 at least one sensor for sensing a vicinity of a vehicle to provide sensed information; 
 at least one processing circuit that is configured to: 
 process the sensed information to provide compact contextual signatures of sensed road users within the vicinity of the vehicle; wherein a compact contextual signature of each of a sensed road user comprises (a) coarse contextual metadata regarding the sensed road user, (b) coarse location information regarding the sensed road user, (c) identifiers of other sensed road users, and (d) coarse situation information; 
 feed the compact contextual signatures to a machine learning process trained to estimate behaviors of road users based on the compact contextual signatures of road users; and 
 predict, by the machine learning process, the behaviors of the sensed road users.